Your achievements

Level 1

0% to

Level 2

Tip /
Sign in

Sign in to Community

to gain points, level up, and earn exciting badges like the new
Bedrock Mission!

Learn more

View all

Sign in to view all badges

Adobe Experience Platform’s Identity Service — How to Solve the Customer Identity Conundrum




Authors: Joseph R. Jones, Priyanka Sharma, Erin Davis, and Jody Arthur

This blog describes how Adobe Experience Platform Identity Service works with identity data to help our customers develop the information-rich profiles they need to engage their customers at a deeper level. Check out the preceding blog about Identity Graph here.

Identity is a core concept when it comes to building engaging and relevant customer experiences. Yet many enterprises struggle with building the rich and accurate 360-degree customer profiles necessary to deliver such experiences. The problems associated with establishing identity for enterprises is either they have an unmanageable amount of data or too little data. However, the primary challenge lies in connecting the data they have in a way that creates a meaningful representation of their customers' identities.

At Adobe, we know that fragmented profiles developed with disconnected data result in failed attempts to deliver relevant, engaging experiences. Adobe Experience Platform Identity Service solves this problem at the root level by consolidating fragmented profiles and providing enterprises unprecedented opportunities to improve their customers' experiences.

Knowing your customers means being with them throughout their journey

In today's digital world, customers leave traces of their journey with every interaction they have with a brand. Some of these traces are clearly attributable to individual customers through authenticated identifiers such as logins, email addresses, and loyalty card numbers. Other bits of information your customers leave exist as anonymous/pseudonymous data such as device IDs and browser cookies.

Both types of data provide some representation of identity, but neither provide the full picture of the affinities and traits of the individual customer. While deterministic data such as authentication events can help us resolve a customer's identity with 100% certainty, it rarely captures the entire customer journey.

For example, a customer may browse a website anonymously on a smartphone and then return to the site later in the day to log in and make a purchase. We can identify the customer that made the purchase. What we don't know is that the same customer looked at 20 different products earlier in the day in an unauthenticated session before making that purchase. The data resulting from these activities appear as two different interactions, and without the ability to connect them, we end up with a fragmented, incomplete profile of that customer.

Identity graphs help solve this problem. An identity graph is a database that stores all identifiers that correlate with individual customers. By combining deterministic data (i.e., that which is attributable to the individual through authentication) with pseudonymous data into an identity graph, we can develop a more accurate representation of individual customers allowing us to develop better, more personalized experiences for them.

Figure 1: Using an identity graph to stitch profile fragments together into a single Adobe Experience Platform Real-Time Unified Profile of an individual customer.Figure 1: Using an identity graph to stitch profile fragments together into a single Adobe Experience Platform Real-Time Unified Profile of an individual customer.

Adobe Experience Platform Identity Service accomplishes this by grouping device IDs into "person clusters" that represent a pseudonymous person. Person clusters are identities based on deterministic data enriched with additional anonymous data associated with an individual through probabilistic matching.

Using an algorithm powered by Adobe Sensei, Identity Service matches device IDs with anonymized data such as obfuscated IP addresses, device types, and browser types to create statistically significant (and thus very reliable) connections between the different devices a customer might use. By matching the different types of data from the customer's online and offline interactions across different devices, we are able to associate those devices as well as other attributes to the same person to create a "person cluster," which provides a much more robust representation of that individual than would be possible with deterministic data alone. Currently, probabilistic matching is only available with Co-op Graph but will become available in Private Graph in 2020.

Person clusters can be used to develop highly personalized experiences for the individual customer and can also help marketers achieve scale through grouping. For example, person clusters can be grouped into household clusters representing people who live under the same roof or into business clusters representing employer-employee relationships. Marketing messages can then be tailored to the larger group.

By connecting the data between all the touchpoints a customer has with a brand, Identity Service helps marketers know their customers more fully through identity graphs, allowing them to nurture customers through their entire journey with more relevant experiences along the way.

High-level view of Adobe Experience Platform Identity Service

Identity is a critical component of everything we do on Adobe Experience Platform and is one of the platform's core services. Identity Service stitches together the identity of the consumer by examining unauthenticated and authenticated interactions with your brand to build an identity graph that connects with the 360-degree customer profile to represent behavior and interests.

In Adobe Experience Platform, the Experience Cloud Identifier (ECID), which is a shared identity namespace used across the platform and Adobe Experience Cloud solutions, provides the basis for customer identity. Within Adobe Experience Platform, Identity Service uses the ECID as the primary ID for devices and the base node for identity graphs, which add the people-centric context that allows our customers to market to real people as opposed to devices.

Figure 2: High-level overview of Adobe Experience Platform’s Identity Service.Figure 2: High-level overview of Adobe Experience Platform’s Identity Service.

Adobe was a pioneer in the use of ECID data to develop customer profiles. To date, Adobe has about 2,300 customers using our ECID service. Now, with the integration of Identity Service into Adobe Experience Platform, we are able to unify the collection of cookie and device data from all Adobe solutions providing our customers the ability to sync their various types of identity data across all the platform services and Adobe solutions they use.

Figure 3: Current and future state of ECID data collection within Adobe Experience Platform.Figure 3: Current and future state of ECID data collection within Adobe Experience Platform.

How Adobe Experience Platform Identity Service works

The process of developing an identity graph in Identity Service involves two main steps. The first step is to collect the necessary identity data using the ECID service, which is Adobe's recommended tool for collecting and transmitting identity data into Adobe Experience Platform. Then all the identity information must be resolved.

Identity resolution first requires deduplication of the identity data collected from various data sources in the graph and then matching the remaining identity data with the individual customers.

Both processes require matching unique identity namespaces with the individual customer to whom they belong. The namespace of an identity provides context by relating the identity value to its system of origin, which is necessary to fully qualify the identity. This is because identities are always defined by context.

Think of a passport or a driver's license. Without context, the information they contain would be a string of numbers with no meaning. We understand their meaning only when they are associated with a passport or driver's license namespace. Therefore, for the purposes of developing an identity graph, identity is comprised of both an ID value and the namespace associated with that value.

Identity Service provides a standard set of identity namespaces, which are available to all Adobe Experience Platform customers. These are namespaces Adobe has researched and found to be common to all or most customers across verticals (e.g. phone numbers and email addresses). Identity Service also provides the ability for our customers to add their own custom namespaces, such as loyalty IDs, that they can use in developing their identity graphs.

Within the Identity Service schema editor, data engineers and architects can perform identity mapping by selecting from both standard and custom namespaces. Data engineers and architects then can put together a fully qualified identity and obtain additional linked identity data from other namespaces to enrich the profile.

Figure 4. Example of an Adobe Experience Platform Real-Time Unified Profile created with Identity Service containing both standard namespaces and custom namespaces linked to the individual customer.Figure 4. Example of an Adobe Experience Platform Real-Time Unified Profile created with Identity Service containing both standard namespaces and custom namespaces linked to the individual customer.

Working with identity graphs using Identity Service integrations within Adobe Experience Platform

Identity Service integrates with a broad set of APIs within Adobe Experience Platform. to powers cross-device features and solutions. Here are a few examples:

  • With Adobe Analytics, you can compress all your unique visits down to the number of people who actually generated those visits. This integration also will also provide Cross-Device Analytics very soon that will allow you to visualize the paths that consumers take across your website and apps and with any device even when they are not authenticated.
  • You can use your identity graphs with Adobe Audience Manager to expand segments to include other devices belonging to the same consumers as devices that natively qualify for the segment.
  • Adobe Target allows you to extend personalization to additional devices with AI-powered automation at scale and to optimize those experiences with A/B and multivariate testing.
  • Identity graphs can also be used with Adobe Ad Cloud for better forecasting and to provide more accurate people-based attribution allowing you to develop messages for people instead of devices.

Deciding what kind of identity graph is right for your business

At Adobe, our goal is to help our customers build the best identities and profiles to meet their business goals. Currently, Adobe Experience Platform Identity Service supports three types of identity graphs:

  • Private Graph
  • Co-op Graph
  • Third-Party (3P) Graphs from LiveRamp and TapAd later in early 2020 with support from Adobe Experience Platform

The type of identity graph that will best serve your business depends primarily on how much data you have to build it. Figure 5 provides a simple decision tree you can use to determine which kind of identity graph you should start with.

Figure 5: Decision tree for determining the type of identity graph to use within Adobe Experience Platform Identity Service.Figure 5: Decision tree for determining the type of identity graph to use within Adobe Experience Platform Identity Service.

Here are the details regarding the three types of identity graphs.

Private graph

Private graph helps Adobe customers connect their online and offline data and provides identity resolution capabilities for use with various components and services on the Adobe Experience Platform.

Private graphs are available only to the enterprises that create them because they draw from an enterprise's proprietary data, such as information included in its CRM system, customer support IDs, and loyalty card information.

Figure 6: The different types of data commonly incorporated into an enterprise’s private graph.Figure 6: The different types of data commonly incorporated into an enterprise’s private graph.

Co-op Graph

When Adobe began developing Identity Service, it started with the Co-op Graph. Over the past three years, it has become one of the richest shared data sets available on the market for the development of customer identities with more than 120 leading brands contributing data, 1.7 billion devices identified and 300 million person clusters in the US and Canada alone.

The Co-op Graph is now growing faster than ever, owing in large part to the rise of Internet-connected devices (i.e. the Internet of Things).

The Co-op Graph doubled in members in 2017 and again in 2018, and Adobe is working to build it ever larger by expanding the reach of Identity Service. Currently, Identity Service covers North America, and Adobe plans to extend it globally to EMEA and APAC regions.

Figure 7: Adobe Experience Platform Co-op Graph provides one of the richest customer data sets for developing customer profiles on the market today.Figure 7: Adobe Experience Platform Co-op Graph provides one of the richest customer data sets for developing customer profiles on the market today.

Here is an example of how Co-op Graph might work for a retail company. The customer journey looks something like this:

  1. The customer visits a website from her computer and logs in, providing authentication data to the system.
  2. An hour later, she visits the website again on her laptop without logging in and browses through several products on the site. Her visit this time is identifiable only as an IP address, which remains anonymous.
  3. While on the train home from work, that same customer visits the site from her phone and follows a deep link to make her purchase. This event is identifiable from the payment information she entered into the system. However, that information may or may not have anything in common with the authentication data she used when she first visited the site because she may have made the purchase as a guest instead of logging in beforehand.

Based on the data received through these interactions, these events are going to appear in the retailer's data set as three separate customer journeys. Fortunately, by virtue of being a co-op member, the retailer will get the benefit of all the links contributed by other members of the co-op that can help determine that all those devices belong to the same person.

Here is another example of a hospitality company. The customer journey might look something like this:

  1. The customer logs into a travel website on his computer to book a flight. As in the first example, the act of logging in provides a deterministic link that allows the travel company to connect that computer with that individual.
  2. Before arriving at the airport, he checks on the flight status from his phone creating a probabilistic link that, with the Adobe Sensei algorithms built into Co-op Graph, allows the travel company to add that person to a person cluster within the co-op.

Currently, Adobe's Co-op Graph is comprised of 76% determinant links and 24% probabilistic links, which provides Adobe Experience Platform customers the rich and reliable identity graphs they need to market to people instead of devices.

Figure 8: Generalized examples of how Adobe Experience Platform Identity Service Co-op Graph works.Figure 8: Generalized examples of how Adobe Experience Platform Identity Service Co-op Graph works.

Third-Party Graph (Coming in 2020)

Adobe Experience Platform is an open ecosystem. We understand that many of our customers are already working with trusted vendor partners who are collecting valuable data for use in developing customer identities.

Third-Party Graph supports device and cookie-level graphs assembled by third-parties, including LiveRamp and TapAd. Depending on the business model or regulatory constraints, some Adobe Experience Platform customers may have relatively little authentication data to use in developing their identity graphs.

Adobe is developing support for third-party graphs as a way to address this challenge by allowing customers to leverage data from other platforms with whom they have relationships. We are currently working with a number of data partners to test this capability and anticipate its release in 2020.

How to derive maximum value from identity graphs, and some important considerations when using them

Adobe Experience Platform Identity Service provides unprecedented opportunities for enterprises to deliver personalized experiences. In this section, we offer a few tips to help Adobe Experience Platform customers get the most value from their data in Identity Service along with some important considerations going forward.

Use A/B testing to derive more value

We recommend using A/B testing to determine which type of graph performs better for the specific use case you have. While A/B testing requires additional work, the value you can derive from it can significantly impact your business.

One way is to set up identical segments to test one type of graph against another. We have seen cases where one segment performs better with Co-op Graph while another segment will perform better with a third party graph, and vice versa.

In an Adobe Audience Manager use case, we could use our identity graphs to expand a segment to include other devices that belong to the same people based on all identities that natively qualify for that segment. In this case, an individual customer's phone may qualify for a segment based on that customer shopping for a laptop with his phone, meaning that the customer's phone now has that trait associated with it. The customer also has an iPad, which doesn't carry that trait. However, because we know that both devices belong to the same person, we can expand the number of devices in that segment. So, a segment with a million devices becomes two million devices through which you can deliver personalized experiences.

A/B testing is particularly valuable in cases where enterprises paying high CPM to reach their target market. In these cases, even a small difference in performance can translate into significant savings. For example, a difference from 95 percent to 105 percent revealed through an A/B test may seem relatively small. But, when evaluated within the context of a CPM costing $200-$300, that lift can have a huge impact on the bottom line of the business.

Create as many custom namespaces as necessary

The more identities you have to work with, the more opportunities you have to create personalized experiences for your customers. In Adobe Experience Platform Identity Service, you have the ability to create as many custom namespaces as you need to. We recommend that enterprises leverage this capability by building as many custom namespaces as possible that are unique to their business. In doing so, they derive greater value from the platform. With Identity Service, you're only limited by your imagination.

Don't be creepy

One of the advantages of having a Private Graph, a Co-op Graph or a Third-Party Graph is that they allow you to use Adobe Target to personalize even on unauthenticated identities. However, there is a right way to personalize, and a wrong way. Here is an example of each:

  • Right Way – Sprint, an Adobe Experience Platform customer, recently ran an ad campaign with a hero graphic that would show either an Android phone or an iPhone based on their customers' preferred devices. When customers are shown the ad on their phones, it's easy for Sprint to show them the correct version. But, if they are visiting the site on a laptop, Sprint has no way to determine which version of the ad would appeal to the customer. However, by including a phone ECID in the person clusters they were targeting with their ad, they were able to achieve a 650% increase in conversion rates.
  • Wrong Way – an example of the wrong way to personalize would be when a customer visits your website, and without ever logging on, you show her an ad that refers to her by name. To a customer who, to her knowledge, hasn't yet provided any identity to your website, this kind of personalization is at best off-putting and at worst, can come off as very creepy.

It's also probably time to think about real-time

At Adobe, we were seeing an emerging global trend in the rise of new privacy controls related to cookie collection. These changes are important because they impact how long a device can hold onto a given cookie, which is commonly referred to as cookie churn.

Cookies represent a snapshot in time. Before regulation, an enterprise could reasonably expect a cookie to persist between 60-90 days. As the use of cookies began to be more regulated, we observed that cookies were persisting only about 30 days. Now with Enhanced Tracking Protection (ETP) by Mozilla and similar efforts by Safari and Chrome, we have begun to see cookies churn in seven days, sometimes less.

This is significant because Co-op Graph and Private Graphs are updated weekly. With cookies churning in seven days, these identity graphs will begin to lose their value because the cookie data informing them will become outdated almost instantly. In a changing climate where enterprises have a higher percentage of short-lived interactions coupled with a much faster cookie churn, it will become necessary to create identity graphs in near real-time to continue to leverage the power they provide for personalization.

Figure 9: Timeline showing the rise in cookie regulation.Figure 9: Timeline showing the rise in cookie regulation.

Follow the Adobe Experience Platform Community Blog for more developer stories and resources, and check out Adobe Developers on Twitter for the latest news and developer products. Sign up here for future Adobe Experience Platform Meetups.

Originally published: Aug 29, 2019